New deployment of 1.7.0-SNAPSHOT affects customer evaluation of regulation rates
Dear colleagues -
Today I deployed an update to the 1.7.0-SNAPSHOT server that fixes Issue #1040 and Issue #1041. Both of these update the handling of tariffs with RegulationRates.
The responseRate field in RegulationRates is now ignored when validating new TariffSpecifications. This field has caused unnecessary confusion, for which we apologize. Previously, the value of the field had to be MINUTES to avoid having RegulationRates ignored. Since we are not currently using this field, and since at least one broker in the 2019 tournament was having its RegulationRates ignored because of this, we decided to just ignore the field when validating new tariff offers.
Customers now discount the potential earnings from RegulationRates based on a rough approximation of how likely the quoted prices are to be cleared in the balancing market. Previously, a tariff with a very high price (paid to the customer) for up-regulation would be highly valued by the customer, and yet that price would never clear and the customer would never be paid. Currently, the discount is computed using a logistic function that produces a multiplier of about 0.95 for a price 3 times the mean market price, 0.5 for a price 4 times the mean price, and about 0.05 for a price 5 times the mean market price. We are still studying the effect of this change and so the exact values may yet change a bit.
As always, please let us know if you have questions or concerns.
Re: New deployment of 1.7.0-SNAPSHOT affects customer evaluation of regulation rates
Dear colleagues -
Today I deployed what I expect to be the last in the 1.7.0-SNAPSHOT series. In case you have tried using the regulation rates recently, the parameters needed a bit of fine-tuning. I wrote a script to walk through all the games in the 2019 tournament to see how the clearing prices in the balancing market compared to the mean wholesale market price from the corresponding boot sessions -- see the image below. I then updated the code to make the customer's discounting behavior closely match these distributions, and I made a minor update in the regulation rates issues by the sample broker to make them more aggressive. If you run a game, you can look in the trace logs to see the values used in the discount computation -- just search for the string "upreg". This log entry comes from the TariffEvaluationHelper in the common module.
I hope to do the 1.7.0 release next week. Unless something needs to be fixed, this will be the version we use in the 2020 tournament.